ai-berkshire: Value Investing Research with Claude Code and Codex

The ai-berkshire MCP Client brings a structured, multi-agent approach to value investing research directly into Claude Code and Codex environments. This Python-based client systematizes the methodologies of Warren Buffett, Charlie Munger, Duan Yongping, and Li Lu, enabling AI agents to conduct professional-grade investment analysis. It transforms a single user with an LLM client into a full investment research team, with reported real-world returns.

Integrating Investment Skills with Claude Code

ai-berkshire is designed to extend Claude Code's capabilities by injecting a suite of investment research skills directly into its global command directory. This integration allows users to invoke specialized research functions as native commands within Claude Code.

To install these skills for Claude Code on macOS or Linux, navigate to the ai-berkshire directory after cloning the repository and execute:

./scripts/install-claude-commands.sh

Windows users leveraging PowerShell or Command Prompt can achieve the same by running:

git clone https://github.com/xbtlin/ai-berkshire.git
cd ai-berkshire
.\scripts\install-claude-commands.bat

Once installed, these skills are available for direct invocation within Claude Code, providing an /investment-research experience.

Connecting to Codex with Prompts and Skills

For Codex users, ai-berkshire offers two distinct integration paths: a set of "skills" and an optional layer of "slash prompts." The skills are designed to provide a similar /investment-research experience to Claude Code.

Codex users on macOS or Linux can install the core skills by running:

./scripts/install-codex-prompts.sh

Windows users can install the Codex skills via PowerShell or Command Prompt:

git clone https://github.com/xbtlin/ai-berkshire.git
cd ai-berkshire
.\scripts\install-codex-skills.bat

Additionally, an optional set of Codex slash prompts can be installed on Windows for further compatibility:

REM Optional: Install Codex slash prompts
.\scripts\install-codex-prompts.bat

The repository maintains three distinct entry points for these integrations: skills/*.md are the source files for Claude Code commands, codex-skills/*/SKILL.md are generated from skills/*.md by scripts/sync-codex-skills.py for Codex skill packages, and codex-prompts/*.md provide the optional Codex slash prompt compatibility layer.

Core Capabilities and Framework Design

ai-berkshire provides a collection of investment research skills that operationalize the methodologies of four renowned value investors. The client leverages multi-agent adversarial analysis to conduct professional-grade research, aiming to redefine the depth and efficiency of investment analysis through AI.

The framework's design is not merely theoretical; it is backed by a real track record. The project reports a +69.29% return for 2024 and +66.38% for 2025, significantly outperforming major indices like the S&P 500, Hang Seng Index, CSI 300, and Nasdaq over the same periods. These figures, derived from a real Futu Securities account, demonstrate the practical application and effectiveness of the AI-driven research framework.

The full framework and comprehensive reports are maintained within the ai-berkshire repository.

ai-berkshire MCP Client: Value Investing with Claude Code & Codex

ai-berkshire: Value Investing Research with Claude Code and Codex

Integrating Investment Skills with Claude Code

Connecting to Codex with Prompts and Skills

Core Capabilities and Framework Design

References